Talking 'bout a revolution

But in the case of LEDs, speaking simplistically, it costs 10 times as much to make a bulb 10 times brighter – because you need 10 times the size of semiconductor chip.

So it s tougher to move up into the higher-wattage applications. It s not just the cost; it s also the heat issues. But we re starting at the bottom and as the efficiencies get better and the costs move down we re going to move up to the higher-wattage applications. At least in my mind, there s no longer any question as to what s going to happen. LEDs are going to take over, but the question is timing.

The reason I say this – and I ve felt this way for the last year or so – is because of the best LED results in laboratories, like the 135 lm/W for a 1 W power LED from Nichia (see related story).

[As an industry], we re over 100 lm/W in the lab, so that s clearly going to happen commercially and that is a high enough efficacy for LEDs to win the battle – in my mind at least.

MH: What is needed for solid-state lighting to become competitive?
GC: Price is a big issue. Certainly on an initial cost basis, a tungsten bulb costs less than a dollar. CFLs, which are very efficient, are a couple of dollars or less. To make a 1000  lumen LED would cost $20 or more. So it s 20 times too expensive in terms of cost per lumen.

Right now, [most] commercial products are in the 80 lm/W range. If that doubles to something like 150 lm/W then you would cut the cost in half and get twice as many lumens out of the same device.

If you can also drive the device at 2 A [instead of the standard 350 mA], you get something like a factor of 5–10 times improvement. That device, which might cost perhaps two or three dollars, will clearly come down in cost in the future, perhaps by a factor of two or more. Then there you are, right in the ballpark with the conventional sources.

We aspire to this LED light bulb, which would be hopefully a 1 mm chip that would give out at least 500 lumens – maybe 1000 lumens if we can get the efficiencies we re talking about – and sell for something like a dollar. So you d have a very cost-competitive source and none of the disadvantages of conventional lamps. It would last a long time, there s no mercury to dispose of, it s fully dimmable, you can change its color and so on.

I think that it s real clear that we will have these sources, even if we can t get to that [performance] level. If we can get to efficiencies of 125 lm/W and several hundred lumens coming out, it s still better than any conventional light source and the cost will still be down in the range where it s pretty competitive.

MH: What does Philips look for in terms of cost and performance?
GC: Lumileds has been doing well as a business and Philips has given us the freedom to do what we re doing. There isn t any specific milestone they re driving towards, where suddenly there will be a switch and they ll go to solid-state lighting. It s going to be an evolution as we work our way up, and we and Philips will do what makes sense in terms of applications – it s very applications-driven.

Speaking for myself, I have an aspiration of getting to 1000 lumens from a 1 mm chip, with something like 150–160 lumens-per-watt performance. At that level you d only have about 3 W of power to dissipate and a light bulb you could handle.

MH: Are any fundamental changes to LED chip design going to be needed?
GC: Bottom line: I think it s more of an evolutionary thing. That s because we re already seeing these lab results as high as 135 lm/W and the internal quantum efficiency (IQE) is pretty high at low currents. The big issue is keeping the IQE high at high current densities.

Lumileds has spent a lot of time and effort trying to understand what causes that drop-off [also known as current droop]. It used to be thought that the cause was these indium-rich clusters, a non-uniformity of indium in the junction that means you lose efficiency when you drive it hard.

Now, we don t believe that s the case. We believe it s an Auger effect (see related Applied Physics Letters paper); and I think that the data are pretty compelling.

By increasing the recombination volumes you can minimize that Auger effect and increase the high-current performance. So that s certainly something we re moving towards. Nobody has a good enough crystal ball to know whether growing these thick recombination structures will require some different form of substrate. I can t answer that, but we re certainly making progress with conventional substrates so far.

I guess I hope that we don t have to do something radically different. But if it turns out that we do need a new substrate, then we ll have to go there.

MH: With chip cost such an important issue, is there a need to move to larger wafers?
GC: It s another case of how and when. If you change the wafer size substantially you have to change all of your equipment in your fab and your growth techniques, so that s a big expense and it has to be justified on return basis. It is going to happen, but I don t really have any detailed feel for the time frame.

Now, the silicon [semiconductor] business goes to larger diameters because they use very large chips. If you have very large chips, you have a lot of edge-loss.

The largest chips that we make routinely are 1 mm, so the edge loss isn t such a big factor. Also, silicon wafers are very flat and beautiful. Typically, LED wafers aren t that perfect, so you have a lot of other issues to deal with. So just improving the yield and the processes at the sizes we are [at] is still a big deal. I don t think you can anticipate a complete transition to larger wafers over the next few years – but it will evolve over time.

MH: LED backlights for LCD TVs and monitors have long been identified as the next major application area for HB-LEDs. Is that taking off?
GC: The flat-panel business is fiercely cost-competitive and there is an entrenched technology for backlights. I guess my opinion is that, as elsewhere, LEDs will ultimately be the light source of choice for that market.

We had originally thought that the improved color gamut was going to be a huge factor and that people would love to have LED backlights because they could get these saturated red, green and blue colors.

Then the color gamut of the existing [fluorescent lamp] technology improved and so it just hasn t happened. So it s going to be a slow evolution, although it seems to me that a solid-state, non-breakable, non-mercury, ultimately lower-power and very cost-effective solution should dominate. But it may take quite a while, because it s just a really tough market.

MH: What would you say is the biggest technological challenge that you face right now?
GC: The biggest thing is the internal quantum efficiency at high drive currents. That s where all the headroom is.

Is a fundamental change going to be required? Well, that remains a question. We re looking at how to do that and if we knew how to do it, we d do it instantly.

We have made improvements, but we need to figure out how to improve by a factor of two at high drive current densities, whatever it takes. Whether it is a breakthrough, tweaking a lot of small things, or a combination of the two is the open question – we ll just have to see.